is a deputy editor and author of the Command Line newsletter.
The day before the launch of GPT-5, OpenAI CEO Sam Altman surprised employees with a message in the company’s Slack.
“As we mentioned a few weeks ago, we have been looking at comp for our technical teams given the movement in the market,” wrote Altman, according to a copy of the message that was shared with me. He announced that OpenAI would give a “special one-time award” to researchers and software engineers in a handful of orgs, including applied engineering, scaling, and safety.
“We very much intend to keep increasing comp as we keep doing better and better as a company,” he wrote. “But we wanted to be transparent about this one since it’s a new thing for us.” He then thanked employees for their “excellent work towards building AGI.”
The bonus amount for each qualifying employee will vary based on role and seniority, according to my sources. The highest payouts will be in the mid, single-digit millions for OpenAI’s most coveted researchers, all of whom already make millions per year. Engineers, meanwhile, are expected to receive bonuses worth hundreds of thousands of dollars on average. The bonuses will be paid out every quarter over the next two years, with the option to receive the money in OpenAI stock, cash, or a mix of both. Roughly 1,000 employees, or about one-third of OpenAI’s full-time workforce, qualify.
At the same time, OpenAI is gearing up to let more employees cash out millions of their vested stock to investors. Altman has privately said he expects they’ll be able to sell at a much higher share price than the $274 price awarded in OpenAI’s funding round earlier this year, which valued the company at $300 billion. Bloomberg reported this week that OpenAI may be valued at $500 billion in the upcoming tender offer.
A spokesperson for OpenAI declined to comment. The Information first reported the news that research and engineering employees were getting a one-time bonus, but Altman’s memo is being reported here for the first time.
OpenAI has never offered such a payout to such a wide swath of employees before. The move highlights how fierce the talent wars for AI talent have become. Altman has lost a string of key researchers to Meta in recent weeks, where one of the creators of ChatGPT, Shengjia Zhao, was just named chief scientist. OpenAI’s chief research officer, Mark Chen, has internally compared Mark Zuckerberg’s poaching spree to a home invasion, while Altman has told employees that “missionaries will beat mercenaries.”
It’s not just Zuckerberg that Altman has to contend with. My sources say that Elon Musk’s xAI has also been making aggressive offers to lure away OpenAI’s top talent. In addition, ex-OpenAI CTO Mira Murati recently poached several of the company’s technical leaders to help start her rival lab, Thinking Machines.
By not giving all of OpenAI the “special” bonus, Altman risks breeding jealousy among his employees who worked nonstop on this week’s launches and were left out. For now, though, everyone seems to be riding high from the buzzy release of GPT-5. After Thursday’s launch event, I’m told that Altman joined employees for a celebratory happy hour at Splash, a sprawling sports bar in the Chase Center near OpenAI’s headquarters in downtown San Francisco.
More on OpenAI
- Anthropic gets dealt a blow. Out of all the companies enthusiastically announcing their support for GPT-5, I took particular note of Cursor’s execs appearing on OpenAI’s livestream to tout that they are making the model their default, replacing Claude. Given that Cursor represents a significant chunk of Anthropic’s revenue, I’m sure this sounded alarm bells internally. It also at least partially explains why Anthropic rushed out an update to Claude that makes it slightly better at coding. Never a dull moment in the AI race!
- Who got to try GPT-5 early? Definitely not me, or seemingly any of my peers in the “legacy” media. Instead, like past model releases, OpenAI identified a handful of independent writers and AI influencers to grant early access to under NDA. (We don’t sign those at The Verge.) It seems like they all got access to GPT-5 about two weeks ago, which makes it all the more impressive that more about the model didn’t leak. If you’re curious to read about the GPT-5 impressions these chosen few have, I recommend the write-ups from the crew at Every, Simon Willison, Ethan Mollick, Matt Schumer, Latent Space, Artificial Analysis, and METR.
Link list
More to click on:
- My interview on Decoder with GitHub CEO Thomas Dohmke.
- Ollama is a new Mac app for running LLMs like GPT-OSS locally.
- Eleven Labs introduced a new model for generating commercially usable music.
- The tech companies that are winning the war for engineering talent.
- “How AI conquered the US economy.”
- Life at AI startups after they get reverse acquihired.
- Please stop: People are trying to call San Francisco’s Mission District “the Arena.”
If you haven’t already, don’t forget to subscribe to The Verge, which includes unlimited access to Command Line and all of our reporting.
As always, I welcome your feedback, especially if you have more chart crimes from today’s GPT-5 launch to show me. You can respond here or ping me securely on Signal.
Thanks for subscribing.
